Every technology shift creates the same tension: the tools that make individuals faster make it harder for organizations to govern. AI coding is no exception, and it’s the reason we built Forge for Cursor.
Available now on the Cursor Marketplace, the new Forge plugin brings enterprise context, governance, and spec-driven development directly into the IDE, helping developers generate software that aligns with enterprise requirements from the start of the development process.
The Problem Nobody Budgets For
Cursor has changed how developers work. Code that took days now takes hours. That kind of speed is exactly why context matters more now, not less. Look at where the tokens in any agentic coding session actually go, and the same pattern shows up across the category, regardless of which tool a team standardizes on.
In a typical 50-turn agentic coding session, roughly 1 million tokens go to input and only 40,000 go to output, a ratio of about 25 to 1. Independent research on multi-agent coding systems backs this up from a different angle: in one study of agentic software engineering tasks, token Consumption is dominated by input tokens, far exceeding both output and reasoning tokens. Two different studies, two different measurements, and the same conclusion either way. Your enterprise is paying premium rates for AI to rediscover context it should have been handed, not for the part where it actually writes code.
The cost isn’t only financial. AI tools that guess at context don’t know your architecture standards, security policies, or compliance obligations. The result is architectural drift: code that works in isolation, fails at scale, and pulls your codebase away from its own specifications one accepted suggestion at a time. That drift compounds fastest in exactly the workflows agentic coding is supposed to accelerate, since nobody’s reviewing every intermediate step the way they would a single pull request.
The Solution: Specifications Where Work Happens
The new Forge plugin on the Cursor Marketplace injects your enterprise specifications directly into the IDE. Instead of the AI burning tokens searching your repository for context, Forge hands it exactly what it needs: your architecture standards, security guardrails, and compliance requirements, scoped precisely to the task in front of it.
This is what the industry has started calling context engineering, giving an AI system the situational awareness to act with precision instead of relying on it to reconstruct that awareness from scratch every time. Forge builds that context once, as a living specification, and Cursor draws on it directly rather than re-deriving it turn by turn. Targeted repository data combined with automated guardrails means code arrives compliant instead of getting remediated after the fact, and the tokens a developer spends go toward producing rather than exploring.
What Enterprises Get
Speed without drift. Developers ship at AI pace while every line stays aligned with enterprise architecture, since the specification travels with the task instead of living in a separate document nobody opens mid-sprint.
Governance without friction. Guardrails run automatically inside the workflow instead of adding a new review step on top of it. Because the AI has business intent and policy in hand before it starts, in place of guessing and correcting after the fact, teams running Forge have seen rework drop by roughly 60%, and delivery time on new development fall by 70 to 85%. The same context-once approach that avoids rebuilding project understanding on every turn has cut token spend by roughly 75% per task in early Forge deployments, freeing teams to choose whichever model fits the job instead of defaulting to whatever’s cheapest per token.
One system of record. Forge also stays the system of record across the toolchain. What happens inside Cursor is visible, auditable, and consistent with everything else running through Forge, so governance doesn’t fragment as teams adopt more AI tools rather than fewer.
Setup in Minutes
Install the plugin from the Cursor Marketplace → connect the Forge MCP → server run your configurations.
The MCP connection is what carries governance into the IDE itself rather than leaving it as a separate check somewhere downstream, so the audit trail starts at the first line of generated code, not after a pull request is opened. Your specifications are live in the IDE the same day.
Flip the Equation
Today the industry accepts roughly 25 tokens of overhead for every token of value produced. We think that math is backwards. Every token spent rediscovering context is a token not spent building your product.
The goal for every engineering leader should be simple: push token spend as close to 100% output as possible. That’s what the Forge and Cursor integration is built to deliver.
Get started: Explore Opsera Forge and install the Forge plugin on the Cursor Marketplace.
